Sync capabilities with upstream
Browse files- build.toml +6 -6
build.toml
CHANGED
|
@@ -11,7 +11,7 @@ src = [
|
|
| 11 |
include = [ "." ]
|
| 12 |
|
| 13 |
[kernel.cutlass_w8a8]
|
| 14 |
-
cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 15 |
src = [
|
| 16 |
"core/math.hpp",
|
| 17 |
"cutlass_w8a8/common.hpp",
|
|
@@ -47,7 +47,7 @@ depends = [ "cutlass_3_6", "torch" ]
|
|
| 47 |
|
| 48 |
[kernel.fp8_common]
|
| 49 |
language = "cuda-hipify"
|
| 50 |
-
cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 51 |
rocm-archs = [ "gfx906", "gfx908", "gfx90a", "gfx940", "gfx941", "gfx942", "gfx1030", "gfx1100", "gfx1101" ]
|
| 52 |
src = [
|
| 53 |
"fp8/amd/hip_float8.h",
|
|
@@ -61,7 +61,7 @@ include = [ "." ]
|
|
| 61 |
depends = [ "torch" ]
|
| 62 |
|
| 63 |
[kernel.fp8_marlin]
|
| 64 |
-
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 65 |
src = [
|
| 66 |
"fp8/fp8_marlin.cu",
|
| 67 |
"gptq_marlin/marlin.cuh",
|
|
@@ -71,7 +71,7 @@ depends = [ "torch" ]
|
|
| 71 |
|
| 72 |
[kernel.int8_common]
|
| 73 |
language = "cuda-hipify"
|
| 74 |
-
cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 75 |
rocm-archs = [ "gfx906", "gfx908", "gfx90a", "gfx940", "gfx941", "gfx942", "gfx1030", "gfx1100", "gfx1101" ]
|
| 76 |
src = [
|
| 77 |
"compressed_tensors/int8_quant_kernels.cu",
|
|
@@ -81,7 +81,7 @@ include = [ "." ]
|
|
| 81 |
depends = [ "torch" ]
|
| 82 |
|
| 83 |
[kernel.gptq_marlin]
|
| 84 |
-
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 85 |
src = [
|
| 86 |
"core/scalar_type.hpp",
|
| 87 |
"gptq_marlin/awq_marlin_repack.cu",
|
|
@@ -94,7 +94,7 @@ include = [ "." ]
|
|
| 94 |
depends = [ "torch" ]
|
| 95 |
|
| 96 |
[kernel.marlin]
|
| 97 |
-
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "
|
| 98 |
src = [
|
| 99 |
"core/scalar_type.hpp",
|
| 100 |
"marlin/dense/common/base.h",
|
|
|
|
| 11 |
include = [ "." ]
|
| 12 |
|
| 13 |
[kernel.cutlass_w8a8]
|
| 14 |
+
cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 15 |
src = [
|
| 16 |
"core/math.hpp",
|
| 17 |
"cutlass_w8a8/common.hpp",
|
|
|
|
| 47 |
|
| 48 |
[kernel.fp8_common]
|
| 49 |
language = "cuda-hipify"
|
| 50 |
+
cuda-capabilities = [ "7.0", "7.2", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 51 |
rocm-archs = [ "gfx906", "gfx908", "gfx90a", "gfx940", "gfx941", "gfx942", "gfx1030", "gfx1100", "gfx1101" ]
|
| 52 |
src = [
|
| 53 |
"fp8/amd/hip_float8.h",
|
|
|
|
| 61 |
depends = [ "torch" ]
|
| 62 |
|
| 63 |
[kernel.fp8_marlin]
|
| 64 |
+
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 65 |
src = [
|
| 66 |
"fp8/fp8_marlin.cu",
|
| 67 |
"gptq_marlin/marlin.cuh",
|
|
|
|
| 71 |
|
| 72 |
[kernel.int8_common]
|
| 73 |
language = "cuda-hipify"
|
| 74 |
+
cuda-capabilities = [ "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 75 |
rocm-archs = [ "gfx906", "gfx908", "gfx90a", "gfx940", "gfx941", "gfx942", "gfx1030", "gfx1100", "gfx1101" ]
|
| 76 |
src = [
|
| 77 |
"compressed_tensors/int8_quant_kernels.cu",
|
|
|
|
| 81 |
depends = [ "torch" ]
|
| 82 |
|
| 83 |
[kernel.gptq_marlin]
|
| 84 |
+
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 85 |
src = [
|
| 86 |
"core/scalar_type.hpp",
|
| 87 |
"gptq_marlin/awq_marlin_repack.cu",
|
|
|
|
| 94 |
depends = [ "torch" ]
|
| 95 |
|
| 96 |
[kernel.marlin]
|
| 97 |
+
cuda-capabilities = [ "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0" ]
|
| 98 |
src = [
|
| 99 |
"core/scalar_type.hpp",
|
| 100 |
"marlin/dense/common/base.h",
|